feat(telemetry): enabled trace setup for evaluations

Author: poshinchen

.github/workflows/integration-test.yml

@@ -62,4 +62,4 @@ jobs:
           python-version: '3.10'
       - name: Install dependencies
         run: |
-          pip install --no-cache-dir hatch
+          pip install --no-cache-dir hatch

pyproject.toml

@@ -19,6 +19,9 @@ dependencies = [
     "strands-agents>=1.0.0",
     "strands-agents-tools>=0.1.0,<1.0.0",
     "typing-extensions>=4.0",
+    "opentelemetry-api>=1.20.0",
+    "opentelemetry-sdk>=1.20.0",
+    "opentelemetry-instrumentation-threading>=0.51b0,<1.00b0",
 ]
 
 [tool.hatch.build.targets.wheel]
@@ -34,18 +37,21 @@ test = [
 
 dev = [
     "hatch>=1.0.0,<2.0.0",
-    "mypy>=1.0",
+    "mypy>=1.15.0,<2.0.0",
     "pre-commit>=3.2.0,<4.2.0",
-    "ruff>=0.4.4,<1.0.0",
+    "ruff>=0.13.0,<0.14.0",
 ]
 
+otel = ["opentelemetry-exporter-otlp-proto-http>=1.30.0,<2.0.0"]
+
 [tool.ruff]
 line-length = 120
 include = ["src/**/*.py", "tests/**/*.py"]
 
 [tool.hatch.envs.hatch-test]
 installer = "uv"
 extra-args = ["-n", "auto", "-vv"]
+features = ["otel"]
 dependencies = [
     "pytest>=8.0.0,<9.0.0",
     "pytest-cov>=7.0.0,<8.0.0",
@@ -79,6 +85,17 @@ prepare = [
 ]
 
 
+[tool.hatch.envs.hatch-static-analysis]
+installer = "uv"
+features = ["otel"]
+dependencies = [
+  "mypy>=1.15.0,<2.0.0",
+  "ruff>=0.13.0,<0.14.0",
+  # Include required pacakge dependencies for mypy
+  "strands-agents-evals @ {root:uri}",
+]
+
+
 [tool.hatch.envs.hatch-static-analysis.scripts]
 format-check = [
     "ruff format --check"
@@ -137,9 +154,9 @@ dependencies = [
 ]
 extra-dependencies = [
     "hatch>=1.0.0,<2.0.0",
-    "mypy>=1.0",
+    "mypy>=1.15.0,<2.0.0",
     "pre-commit>=3.2.0,<4.2.0",
-    "ruff>=0.4.4,<1.0.0",
+    "ruff>=0.13.0,<0.14.0",
 ]
 
 [tool.coverage.run]

src/examples/agents_as_tools.py

@@ -2,6 +2,7 @@
 import datetime
 
 from strands import Agent, tool
+
 from strands_evals import Case, Dataset
 from strands_evals.evaluators import InteractionsEvaluator, TrajectoryEvaluator
 from strands_evals.extractors import tools_use_extractor

src/examples/bank_tools_trajectory.py

@@ -2,6 +2,7 @@
 import datetime
 
 from strands import Agent, tool
+
 from strands_evals import Case, Dataset
 from strands_evals.evaluators import TrajectoryEvaluator
 from strands_evals.extractors import tools_use_extractor

src/examples/dataset_generator/simple_dataset.py

@@ -1,6 +1,7 @@
 import asyncio
 
 from strands import Agent
+
 from strands_evals.evaluators.output_evaluator import OutputEvaluator
 from strands_evals.generators.dataset_generator import DatasetGenerator
 

src/examples/dataset_generator/trajectory_dataset.py

@@ -1,6 +1,7 @@
 import asyncio
 
 from strands import Agent, tool
+
 from strands_evals.evaluators import TrajectoryEvaluator
 from strands_evals.extractors import tools_use_extractor
 from strands_evals.generators import DatasetGenerator

src/examples/evaluate_graph.py

@@ -3,6 +3,7 @@
 
 from strands import Agent
 from strands.multiagent import GraphBuilder
+
 from strands_evals import Case, Dataset
 from strands_evals.evaluators import InteractionsEvaluator, TrajectoryEvaluator
 from strands_evals.extractors import graph_extractor

src/examples/evaluate_swarm.py

@@ -3,6 +3,7 @@
 
 from strands import Agent
 from strands.multiagent import Swarm
+
 from strands_evals import Case, Dataset
 from strands_evals.evaluators import InteractionsEvaluator, TrajectoryEvaluator
 from strands_evals.extractors import swarm_extractor

src/examples/multi_shots.py

@@ -2,6 +2,7 @@
 import datetime
 
 from strands import Agent
+
 from strands_evals import Case, Dataset
 from strands_evals.evaluators import InteractionsEvaluator
 from strands_evals.types import Interaction, TaskOutput

src/examples/safety_judge_output.py

@@ -2,6 +2,7 @@
 import datetime
 
 from strands import Agent
+
 from strands_evals import Case, Dataset
 from strands_evals.evaluators import OutputEvaluator